x
Optical/IP

Dr Roberts: The Internet's in Trouble

It seems every few years, someone predicts doom for the Internet. It's Larry Roberts's turn.

According to Roberts -- well known as a team member behind Arpanet, the project that eventually grew into today's Internet -- it's not technology that has the Internet in trouble, but economics. Specifically, router prices are going to have a crippling effect, he believes.

Roberts's prediction appears today on Light Reading's brainier cousin site, Internet Evolution.

His latest concern is that, while bandwidth requirements continue to climb, router prices aren't falling at a similar rate. "The cost of Internet capacity would therefore double every three years without some key new innovation," he writes. "The economy could not support this for very long."

Those who have followed Roberts's career won't be surprised to hear his answer: flow-based routing, the technology he pursued with startup Caspian Networks and with his current venture, Anagran Inc. (See Flow-Based Networking and Anagran Promises TLC for TCP.)

His complete essay may be found here.

— Craig Matsumoto, West Coast Editor, Light Reading

gocowboys 12/5/2012 | 3:00:03 PM
re: Dr Roberts: The Internet's in Trouble This is just blatant marketing. They are developing a completely new technology requiring network forklift upgrades. Clearly, they want to create a paranoia and buzz that merits this type of change.

For anyone interested, I am attaching a Larry Roberts article vintage 2003 on flow routing:

http://www.packet.cc/files/Flo...

I just love little companies that are trying to boil the ocean. :-)




tsat 12/5/2012 | 3:00:04 PM
re: Dr Roberts: The Internet's in Trouble
Good point. The fact is that routers work. They might not be that sexy anymore, but they are the proven foundation of the Internet.

Its like saying "Man, that Intel microprocessor technology is only incrementally improving, we need a radically different solution!"

Unlikely to happen.

-tsat
thausken 12/5/2012 | 3:00:05 PM
re: Dr Roberts: The Internet's in Trouble I've posted the following message also with the other article:
http://www.lightreading.com/do...

Let's start with fundamentals. What is the problem that needs solving? If there is demand for more capacity (more users? more traffic per user?), then that should mean more money to providers to provide those services, and more money to box companies to make routers. In other words, Internet capacity is not a fixed resource, and the total cost should scale with the total demand. That is a good thing.

If the growth in demand doesn't grow revenues, then the implicit suggestion is that somehow the users shouldn't or won't pay more for getting more capacity. That can only happen if engineers find cheaper solutions or the providers and box companies cut their margins.

I just don't see the connection from the status quo to a collapse of the Internet. If someone can explain that, please do.

Tom Hausken
Strategies Unlimited
IPforEverything 12/5/2012 | 3:00:06 PM
re: Dr Roberts: The Internet's in Trouble Actually we can displace a lot of the requirements for more routers. Many networks use MPLS at the core. MPLS switches can be more cost effective because because they have to make simpler decisions. Much of the additional traffic is video which can be cached at the edge avoiding most of the routing problems. Also P2P systems tend to reduce the the number of router hops required to get the data. So the fact that the cost electronics is not decreasing as fast as Internet traffic is rising is mitigated by other technologies that are being used to solve the
the cost problem.


OldPOTS 12/5/2012 | 3:00:07 PM
re: Dr Roberts: The Internet's in Trouble "centralized arch will be easy for operation and service, the service will becoming the P2P or P2MP oriented, the control point can be actas the proxy of the P2P service, the end user will be easily to share the information with a high speed."

I understood that the reason we went to IP & Ether was to avoid centralized control of P2P and P2MP service?

And then how is Ether cheaper than IP if you have to buy switches to replace the routers already purchased?

Just wondering - OP
douaibei 12/5/2012 | 3:00:07 PM
re: Dr Roberts: The Internet's in Trouble The reason we have router is first, too much different protocol, second, so many interfaces
if everything is finally migrated to the IP and ethernet, the network architecture will be less rely on the router.

the centrallized forwarding will be the key, as the all of the intelligent stuff will be done at a control point, and the some of the service can be done by the serve system not the router.

centralized arch will be easy for operation and service, the service will becoming the P2P or P2MP oriented, the control point can be actas the proxy of the P2P service, the end user will be easily to share the information with a high speed.

the future will depend on the optical netowrk and the Ethernet switching capability, the rest of the intelligence should be done by the service or cluster system, which has high performance, higher flexibility and lower cost.

the interesting topic will be how the X86 or similar technology will change the network as service evolves.
Pete Baldwin 12/5/2012 | 3:00:07 PM
re: Dr Roberts: The Internet's in Trouble PBB-TE ... should have known that would get brought up sooner or later! They didn't win a Leading Light for nothing.

http://www.lightreading.com/do...

Continuing with the theme of better economics, we've posted a Black Swan analysis from Scott Raynovich. Good stuff:

http://www.lightreading.com/do...
Pete Baldwin 12/5/2012 | 3:00:07 PM
re: Dr Roberts: The Internet's in Trouble >What is interesting more so is that they are focused on a market that is dominated by a few... and lets face it charge way over the top.

That comment's gotten a "2" rating so far, but I think there's some credence here. As the router market expands, the economics should loosen up.
macster 12/5/2012 | 3:00:08 PM
re: Dr Roberts: The Internet's in Trouble Grunt,

The 1/3 argument comes about using the case scenario of a 'regular router' running at 30% capacity (due to over-provisioning). A 'flow router' can run at 90% due to the 'flow-networking' approach employed (apparently).

I suggest you read up on flow networking and related issues such as TCP synchronisation, etc.

jepovic 12/5/2012 | 3:00:08 PM
re: Dr Roberts: The Internet's in Trouble Sure, that sounds like a more realistic solution: Replace some or many of the routers with switches. Won't make the routers any cheaper, but it will make the Internet cheaper.

As far as Dr Roberts, I don't understand how his model would lower the cost of routers. Sure, there is probably room for cutting router prices, but that's a whole different question. I have more hope in vendors like Alcatel and Huawei copying Cisco functionality on a much lower price level.
jhawkins 12/5/2012 | 3:00:10 PM
re: Dr Roberts: The Internet's in Trouble We've been pointing out that routers are expensive for some time now. Glad to see a rationale for how this fact will impact the net going forward.

Of course, we think Ethernet switches are the answer to that particular conundrum (having been proven to be less costly for several decades now). Getting rid of routers... now that's a price curve I'd like to ride for a while. How can that be a bad thing?

And hey, if those switches are PBB-TE enabled, you don't to throw the baby out w/ the bathwater!

<flame shields="" up="">...

john
</flame>
grunt 12/5/2012 | 3:00:11 PM
re: Dr Roberts: The Internet's in Trouble I just don't get this argument - it seems like pure marketing .

Why would a flow router be 1/3 the cost of a regular router? just because it only has to lookup 1/3rd the IP addresses in a route table? DoesnGăÍt a flow router still have to look up the flow for each packet? Even if it only has to look up 1/3rd compared to a regular router why would this portion of the functionality of a router drop the cost of the entire product to a 1/3rd?

Also the benefit of throttling at the edge - I think most or all edge routers can do this.

Also the ability of routers to have low latency flows in the face of all other traffic has been proven I think even by light reading testing for both core and edge products. In fact I think this general problem probably already has many commercially available options (MPLS, setting QoS in IP, and now PBT...)

So at best I can see his proposed flow router being cheaper (although Im not sure why t would be lower cost but can easily be lower price), and by his own admission cheaper is only a step function, not a long term solution to differing rates of growth.

It seems that many times we read about these problems which need to be solved, but in fact often there is existing technology commercially available often by leading vendors available to solve them, but they are not yet deployed. The real issues / constraints seem to often be operational. Maybe Avici is on to somethingGă¬.
Pete Baldwin 12/5/2012 | 3:00:11 PM
re: Dr Roberts: The Internet's in Trouble I was wondering if someone would say that. :)

What I wonder about his argument is whether router prices might drop on a steeper curve in the future. Off-the-shelf chips are there to build routers from, and black-magic arts like BGP are becoming known to a wider and wider circle of programmers (*and* off-the-shelf software is available too).
bollocks187 12/5/2012 | 3:00:11 PM
re: Dr Roberts: The Internet's in Trouble What is interesting more so is that they are focused on a market that is dominated by a few... and lets face it charge way over the top. The Chinese vendors have products that work just as well and at half the price.

Dr. Roberts is changing the marketing game - kudos to him and his team.
dbostan 12/5/2012 | 3:00:11 PM
re: Dr Roberts: The Internet's in Trouble Nice sales pitch!
Buying from his company would solve the internet problem, isn't it?
grunt 12/5/2012 | 3:00:11 PM
re: Dr Roberts: The Internet's in Trouble I just don't get this argument - it seems like pure marketing .

Why would a flow router be 1/3 the cost of a regular router? just because it only has to lookup 1/3rd the IP addresses in a route table? DoesnGăÍt a flow router still have to look up the flow for each packet? Even if it only has to look up 1/3rd compared to a regular router why would this portion of the functionality of a router drop the cost of the entire product to a 1/3rd?

Also the benefit of throttling at the edge - I think most or all edge routers can do this.

Also the ability of routers to have low latency flows in the face of all other traffic has been proven I think even by light reading testing for both core and edge products. In fact I think this general problem probably already has many commercially available options (MPLS, setting QoS in IP, and now PBT...)

So at best I can see his proposed flow router being cheaper (although Im not sure why t would be lower cost but can easily be lower price), and by his own admission cheaper is only a step function, not a long term solution to differing rates of growth.

It seems that many times we read about these problems which need to be solved, but in fact often there is existing technology commercially available often by leading vendors available to solve them, but they are not yet deployed. The real issues / constraints seem to often be operational. Maybe Avici is on to somethingGă¬.
tsat 12/5/2012 | 3:00:12 PM
re: Dr Roberts: The Internet's in Trouble
Yes, the serious problem with the Internet is that it relies on products that my company does not sell.

-tsat
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE